retrieval of aerosol optical properties and estimation of aerosol forcing ...

RETRIEVAL OF AEROSOL OPTICAL PROPERTIES AND ESTIMATION OF AEROSOL FORCING BASED ON MULTI-SPECTRAL SUN PHOTOMETER OBSERVATIONS *

Krzysztof M. Markowicz , Aleksandra E. Kardaś Institute of Geophysics, Warsaw University, Poland

aerosol forcing. Precise information about these

ABSTRACT

quantities enables computing based on radiative *

We discuss an inversion method to retrieve

transfer models the reduction of solar radiation at

aerosol optical properties based on direct and

the Earth’s surface and change of planetary

diffuse multi-spectral observation of solar radiation

albedo at the top of the atmosphere. Models

at the Earth’s surface. These radiation quantities

calculations require information about spectral

are measured by a sun photometer with CCD

aerosol optical properties and its variability with

spectrometer.

the

altitude. Much of the recent work has been

instrument detector is between 350 and 1100 nm

devoted to improve observational networks such

and includes 255 channels. This instrument is

as the Aerosol Robotic Network (AERONET)

operated with two modes; one with active solar

(Holben et. al, 1998), a European Aerosol

tracking, which allows measuring direct solar

Research Lidar Network (ERLINET), and the

radiation and the second with almucantar or

micro-pulse lidar network (MPLNET). In spite of

principle plane scans, which are used to measure

this fact there is no technique to measure remotely

spectral sky radiance. We retrieve following

aerosol optical and physical properties required by

parameters: spectral aerosol optical thickness,

radiative transfer models. For instance lidar

columnar single scattering albedo, asymmetry

observation is usually limited to measure profiles

parameter and total water vapor. Finally basing on

of extinction and backscatter coefficients only. In

these

addition, aerosol lidar retrieval requires strong

The

parameters

spectral

and

range

radiative

of

transfer

calculations the aerosol direct radiative forcing

assumptions,

was estimated.

uncertainties (Welton et al., 2002). The

1. INTRODUCTION

sun

which and

sky

introduce radiance

significant observation

(AERONET) is used to retrieve vertically averaged quantities such as single scattering albedo (SSA),

Knowledge of aerosol optical properties plays

scattering phase function. However these aerosol

an important role in estimation of the direct

optical properties are limited only to a few spectral

*

Corresponding author address: Krzysztof M. Markowicz, Institute of Geophysics, Warsaw University, Pasteura 7, 02093 Warsaw, Poland

data points. Wavelength variability of the aerosol optical properties is important for the aerosol

radiative forcing but also for the identification of

to

aerosol type.

measurement.

In this paper we discuss new observational

block

two

channels

and

perform

dark

Both tubes limiting the FOVs were mounted

technique to measure and retrieve multi-spectral

on two-axis sun tracker.

aerosol optical properties. We present results

follow up the sun with precision about 0.01 and it

obtained during SAWA experiment, which took

performs sky scans (almucantar, principle or

place in Warsaw, Poland in 2005.

others). The MSSP instrument is controlled by

This tracker allows to o

MatLab software by USB2 and RS232 ports. The MSSP is calibrated according to the

2. AUTOMATIC SUN AND SKY SCANNING

Langley method. We are using this technique for

MULTI-SPECTRAL SUN PHOTOMETER

the SUN channel. For the SKY channel the In this section we describe Multi-Spectral Sun

Langley method cannot be applied because direct

Photometer (MSSP), which is used to measure

solar radiation saturates the MMS1 spectrometer.

spectral: direct and diffuse solar radiance. The

In order to calibrate SKY channel we compare it

main

with the SUN one in the region of sun aureole.

part

of

this

instrument

is

a

CCD

spectrometer with 256 channels. We use the MMS1 UV-VIS Zeiss spectrometer with the

3. RETRIEVAL OF SPECTRAL AEROSOL

spectral range between 320 and 1100 nm.

OPTICAL PROPERTIES

Resolution of this detector is about 3 nm and wavelength accuracy is about 0.2 nm. The

The AOT measurements have been made for

measurement time (integration time) can be

many years with two general techniques. One

adjusted from 1.5 to 6500 ms. Noise of this

approach uses a narrow field of view radiometer

detector is small and does not exceed 3 digital

pointed directly at the sun (Volz, 1959) and the

number (DN). Although, the dynamic range of the

second one uses a shadow band radiometer that

MMS1 spectrometer is determined by 15 Bit A/D

measures the total and diffuse solar radiations.

converter (33 768 DN) it is too small to measure

The MSSP instrument uses the first approach,

direct solar and sky radiance. Therefore we

where the AOT is calculated from Beer-Lambert

measure these radiance components with different

law. For this purpose we use observation of the

field of view (FOV). The direct solar radiance is

direct solar radiance. For significant part of

o

o

detector pixels simple Beer-Lambert relationship

FOV. To reduce the direct solar radiance (in the

cannot be applied because of oxygen, ozone, and

case of the smallest FOV) we use 10% of

water vapor absorption bands. Correction for

transmission gray filter. In addition the integration

these

measured by 1 FOV and sky radiance with 2

3

gasses

requires

radiative

transfer

larger if

calculations of the spectral transmittance as a

compared to direct solar radiance. We use two

function of solar zenith angle and total amount of

fiber shutters to block one channel (sun or sky) or

gases in the vertical column.

time for sky radiance is about 10

Based on the differential absorption technique

negligible.

the total water vapor is estimated. For this purpose we use the water vapor absorption band at 936 nm and radiative transfer model. We use MODTRAN, ver. 4.1, which was run in the transmittance

mode

with

20

cm

-1

spectral

resolutions. Retrieval of the SSA and scattering phase function or the asymmetry parameter is more complicated. Distribution of downward diffuse sky radiance for almucantar scan can be described by I(Θ, λ) = Fo e − m o τ m o τ[ωP(Θ) + MS(...)]

(1)

where, Θ is scattering angle, mo is the air mass factor, τ total optical thickness, Fo is the direct flux at the top of the atmosphere, and MS(…) term describes multiple scattering. Thus ratio of diffuse to direct solar radiance is a function of aerosol and molecular optical properties R (Θ, λ) =

I(Θ, λ) = m o τ[ωP(Θ) + MS(...)] . Fdir (λ)

(2)

Fig.1 Calculated sky radiance ratio for different aerosol optical properties based on single scattering approximation (opened points) and DISORT included multi-scattering (dotted points). Blue pointes correspond to the sky radiance ratio for two aerosol optical models. Both models have the same asymmetry parameter (0.75) but different value of the SSA (1.0 for the first model and 0.9 for second). Similar for red points but in this case the SSA is constant and set to value 0.9 and the asymmetry parameter is equal 0.6 for the first model and 0.7 for second model.

In the case of the single scattering approximation angular variability of the diffuse to direct ratio R(Θ,λ) is only function of the scattering phase function. Fig. 1 shows calculated sky radiance ratio for different aerosol optical properties as a function of the scattering angle (for almucantar scan). Sky radiance changes significantly larger due to variability of the asymmetry parameter (red lines) than for the SSA (blue lines). Note the model calculation was performed for the AOT of 0.2 and for this value difference between the single scattering

approximation

(open

circles)

and

multiple scattering model (dotted squares) is not

In order to calculate the distribution of sky radiance we use the same radiative transfer model. Modtran uses DISORT solver, which enables multi-scattering calculation up to 16 radiance streams.

As the input data we use

standard atmospheric profiles, however we scale the input profiles: the total water vapor and total ozone according to the observation. MODTRAN requires information about spectral aerosol optical properties

such

as

coefficient,

the

SSA

the and

aerosol the

extinction asymmetry

parameter. These spectral aerosol quantities are constant with altitude (MODTRAN allows to set four different aerosol models). However, profiles of the aerosol extinction coefficient at 550 nm can by applied independently. In this study we assume

exponential decreasing of the aerosol extinction

springtime. Therefore, we decided to carry out this

coefficient with the altitude. The assumption for

experiment between end of March and May.

typical profile, for most of the atmospheric

To the objectives we estimated the AOT of

conditions does not influence critically the retrieval

spherical and nonspherical particles using 3

quantities (Dubovik et al., 2000).

wavelengths lidar, MSSP, MICROTOPS, and Multi

In this paper we assumed values of spectral

Filter Rotating Shadow Band Radiometer. The

surface albedo concerning the surface type.

contribution of the nonspherical AOT to the total

Because surface albedo has significant influence

AOT was determined from analysis of lidar

on the multiple scattering (especially over the

depolarization at 532 nm. The comparison of the

surface with large albedo) in the future we will

observed radiative fluxes during the days with and

attempt to measure spectral albedo by the MSSP

without Saharan dust events gives an evidence of

instrument.

the aerosol forcing caused by nonspherical

Based on the MODTRAN we developed large

aerosol.

5D look-up table. This table includes following

The mean reduction (during SAWA) of the

independent variables: AOT, SSA, asymmetry

solar radiation at the surface due to aerosol is

parameter, solar zenith angle and relative azimuth

about 6% and the mean aerosol forcing is about

angle. By minimizing the difference between

-18 Wm . The aerosol forcing efficiency was

modeled and observed sky almucantar radiances

estimated between -70 and -77 Wm

at each wavelength we retrieve the SSA and the

values are characteristic for the moderately

asymmetry parameter.

absorbing aerosol. The change of aerosol forcing

-2

-2

and these

The aerosol radiative forcing is estimated from

due to nonspherical particles is small however the

the aerosol optical properties and radiative

difference of radiative forcing between spherical

transfer calculation of total downward and upward

and nonspherical particles strongly depends on

fluxes. For this purpose the aerosol optical

the aerosol size and the aerosol shape. The

quantities are spectrally extended up to 4 µm. The

results of this study are discussed by Markowicz

Angstrom relationship for the AOT and the SSA is

et al. (2005).

assumed.

Fig. 2 presents spectral variability of the AOT (Fig. 2a) and the SSA (Fig. 2b) obtained using the

4. RESULTS AND CONCLUSION

MSSP observation as described above in section 3. Different slope of the AOT in the visible range

In this section we present the results obtained

indicates different particles effective size. The

during SAWA experiment, which took place in

mean (defined for the entire measured spectrum)

Warsaw, Poland in April and in May 2005. The

Angstrom exponent in April 20 is 1.17 while in May

main goal of this campaign was to estimate the

20 is 0.78. The wavelength variability of the SSA

aerosol forcing of the nonspherical Saharan dust.

for these two days is similar, but values of the

Saharan dust over Central Part of Europe can be

SSA are significantly different.

observed several times a year, especially during

This method makes possible analysis of the

Holben, B. N., T. F. Eck, I. Slutsker, D.

spectral variability of the aerosol optical properties

Tanré, J. P. Buis, A. Setzer, E. F. Vermote,

which are function of the particles size distribution,

J. A. Reagan, Y. J. Kaufman, T. Nakajima, F.

shape, and refractive index. Thus our retrieved

Lavenu,

parameters may be used in inversion methods in

AERONET – A federated instrument network and

order to estimate the aerosol size distribution.

data archive for aerosol characterization. Remote

Uncertainties of the retrieval of the SSA and

I.

Jankowiak,

A.

Smirnov,

1998:

Sensing of Environment, 66(1), 1-16.

the scattering phase function or the asymmetry parameters are reduced due to observational

Markowicz, K. M, A.E. Kardas, C. Hochherz, K.

technique. Estimation of these quantities requires

Stelmaszczyk, A.Rozwadowska, T. Zielinski, G.

only relative calibration between SUN and SKY

Karasinski,

channels. Absolute calibration is needed only for

Malinowski, T. Stacewicz, L. Woeste, 2005:

the AOT retrieval. However, the AOT includes

Observation of optical properties and radiative

usually smaller errors than others optical

forcing of nonspherical particles over Poland,

properties due to simpler methodology.

Accent symposium, Sep 2005 – Aerosol: air

J.

Remiszewska,

M.

Witek,

S.

quality and climate. Welton, E. J., J. R. Campbell, 2002: Micro-pulse Lidar Signals: Uncertainty Analysis, J. Atmos. Oceanic Technol., 19, 2089-2094. Volz,

F.

E.,

photoement

1959:

Photometer

zurspektralen

mit

Selen-

Messung

der

Sonnenstrahlung und zer Bestimmung der Wallenlangenabhangigkeit

der

Dunsttrubung

(in

German). Arch. Me-teor. Geophys. Bioklimatol., Fig. 2 Spectral variability of the AOT (a) and the SSA (b) retrieved from observation performed in April 20 and May 20, 2005.

B10, 100–131. 6. ACKNOWLEDGMENTS

5. BIBLIOGRAPHY This

research

was

supported

by

Grant

Dubovik, O., M. D. King, 2000: A flexible

2P04D06927 of Polish Committee for Scientific

inversion algorithm for retrieval of aerosol optical

Research. We would like to thank the Foundation

properties

radiance

for Polish Science for the support. Conference

measurements, J. Geophys. Res., 105, D16,

presentation is supported by EVK2-CT2002-

20673-20,696.

80010-CESSAR EC V-th Framework Program

from

Sun

and

sky

project.